通过多目标贝叶斯优化实现成本感知的上行链路MPQUIC调度

中文摘要 AI 辅助

多路径QUIC（MPQUIC）能在Wi-Fi和LTE等异构接入网络上同时进行上行链路传输，提升可靠性和性能。但积极利用LTE会增加运营成本，导致上传延迟和蜂窝使用之间存在内在权衡。现有MPQUIC调度器通常优化单一性能目标，未明确支持成本感知操作。本文将上行链路MPQUIC调度制定为联合考虑最大上传完成时间和总LTE使用的多目标优化问题。提出基于贝叶斯优化的框架，将MPQUIC系统视为黑盒，系统探索概率路径选择配置以发现帕累托有效操作点。实验表明该方法能刻画宽延迟-成本区域，识别出在上传时间可控增加情况下能大幅节省LTE（高达80%）的配置。结果还表明，在更高竞争水平下，系统多目标探索在成本感知异构上行链路部署中比固定策略调度器更具灵活性。

英文摘要

Multipath QUIC (MPQUIC) enables simultaneous uplink transmission over heterogeneous access networks such as Wi-Fi and LTE, improving reliability and performance. However, aggressive LTE utilization increases operational cost, creating an inherent trade-off between upload delay and cellular usage. Existing MPQUIC schedulers typically optimize a single performance objective and operate at fixed points within this trade-off space, without explicitly supporting cost-aware operation. This paper formulates uplink MPQUIC scheduling as a multi-objective optimization problem that jointly considers maximum upload completion time and total LTE usage. We propose a Bayesian Optimization-based framework that treats the MPQUIC system as a black box and systematically explores probabilistic path selection configurations to uncover Pareto-efficient operating points. Rather than committing to a predefined scheduling policy, the framework exposes a spectrum of delay--cost trade-offs without modifying protocol internals. Experiments conducted using the Mininet-WiFi emulator show that the proposed approach characterizes a wide delay--cost region and identifies configurations that achieve substantial LTE savings (up to 80%) with controlled increases in upload time. The results further indicate that, under higher contention levels, systematic multi-objective exploration provides increased flexibility compared to fixed-policy schedulers in cost-aware heterogeneous uplink deployments.

补充信息

备注

The paper was submitted by a co-author without the consent and approval of the first author and other co-authors